Image Classification on Loc-MNIST 1:1
86.79AccuracyVP (white-box)
Evaluation results
| Method | Links | |
|---|---|---|
| VP (white-box)Backbone=CLIP ViT-B/16, Few-shot samples=16, Access=White-box2024.07 | 86.79 | |
| BlackVIPBackbone=CLIP ViT-B/16, Encoder=vit-mae-base, Few-shot samples=162024.07 | 69.08 | |
| BlackVIP-SEBackbone=CLIP ViT-B/16, Encoder=PCA or kernel PCA, Few-shot samples=162024.07 | 62.92 | |
| BARBackbone=CLIP ViT-B/16, Few-shot samples=162024.07 | 33.98 | |
| ZSBackbone=CLIP ViT-B/16, Prompt=a photo of {classname}2024.07 | 29.7 | |
| VP w/ SPSA-GCBackbone=CLIP ViT-B/16, Few-shot samples=16, Optimizer=SPSA-GC2024.07 | 16.21 |